org.nd4j.linalg.factory.Nd4j.createArrayFromShapeBuffer()方法的使用及代码示例

x33g5p2x  于2022-01-24 转载在 其他  
字(9.5k)|赞(0)|评价(0)|浏览(101)

本文整理了Java中org.nd4j.linalg.factory.Nd4j.createArrayFromShapeBuffer()方法的一些代码示例，展示了Nd4j.createArrayFromShapeBuffer()的具体用法。这些代码示例主要来源于Github/Stackoverflow/Maven等平台，是从一些精选项目中提取出来的代码，具有较强的参考意义，能在一定程度帮忙到你。Nd4j.createArrayFromShapeBuffer()方法的具体详情如下：
包路径：org.nd4j.linalg.factory.Nd4j
类名称：Nd4j
方法名：createArrayFromShapeBuffer

Nd4j.createArrayFromShapeBuffer介绍

暂无

代码示例

代码示例来源：origin: deeplearning4j/nd4j

  1. public abstract DataBuffer decompress(DataBuffer buffer);

代码示例来源：origin: deeplearning4j/nd4j

  1. /**
  2.  * Read in an ndarray from a data input stream
  3.  *
  4.  * @param dis the data input stream to read from
  5.  * @return the ndarray
  6.  * @throws IOException
  7.  */
  8. public static INDArray read(DataInputStream dis) throws IOException {
  9.   DataBuffer shapeInformation = Nd4j.createBufferDetached(new long[1], DataBuffer.Type.LONG);
  10.   shapeInformation.read(dis);
  11.   int length = Shape.length(shapeInformation);
  12.   DataBuffer data = CompressedDataBuffer.readUnknown(dis, length);
  13.   return createArrayFromShapeBuffer(data, shapeInformation);
  14. }

代码示例来源：origin: deeplearning4j/nd4j

  1. INDArray arr = Nd4j.createArrayFromShapeBuffer(buff.dup(), shapeBuff.dup());
  2.   return Pair.of(arr, byteBuffer);
  3. } else {
  4.       new CompressedDataBuffer(byteBufferPointer, compressionDescriptor);
  5.   INDArray arr = Nd4j.createArrayFromShapeBuffer(compressedDataBuffer.dup(), shapeBuff.dup());

代码示例来源：origin: deeplearning4j/nd4j

  1. @Override
  2. public INDArray bitmapEncode(INDArray indArray, double threshold) {
  3.   DataBuffer buffer = Nd4j.getDataBufferFactory().createInt(indArray.length() / 16 + 5);
  4.   INDArray ret = Nd4j.createArrayFromShapeBuffer(buffer, indArray.shapeInfoDataBuffer());
  5.   bitmapEncode(indArray, ret, threshold);
  6.   return ret;
  7. }

代码示例来源：origin: deeplearning4j/nd4j

  1. /**
  2.  * This method creates compressed INDArray from Java float array, skipping usual INDArray instantiation routines
  3.  *
  4.  * @param data
  5.  * @param shape
  6.  * @param order
  7.  * @return
  8.  */
  9. @Override
  10. public INDArray compress(float[] data, int[] shape, char order) {
  11.   FloatPointer pointer = new FloatPointer(data);
  12.   DataBuffer shapeInfo = Nd4j.getShapeInfoProvider().createShapeInformation(shape, order).getFirst();
  13.   DataBuffer buffer = compressPointer(DataBuffer.TypeEx.FLOAT, pointer, data.length, 4);
  14.   return Nd4j.createArrayFromShapeBuffer(buffer, shapeInfo);
  15. }

代码示例来源：origin: deeplearning4j/nd4j

  1. /**
  2.  * This method creates compressed INDArray from Java double array, skipping usual INDArray instantiation routines
  3.  *
  4.  * @param data
  5.  * @param shape
  6.  * @param order
  7.  * @return
  8.  */
  9. @Override
  10. public INDArray compress(double[] data, int[] shape, char order) {
  11.   DoublePointer pointer = new DoublePointer(data);
  12.   DataBuffer shapeInfo = Nd4j.getShapeInfoProvider().createShapeInformation(shape, order).getFirst();
  13.   DataBuffer buffer = compressPointer(DataBuffer.TypeEx.DOUBLE, pointer, data.length, 8);
  14.   return Nd4j.createArrayFromShapeBuffer(buffer, shapeInfo);
  15. }

代码示例来源：origin: deeplearning4j/nd4j

  1. @Override
  2. public INDArray dup() {
  3.   WorkspaceUtils.assertValidArray(this, "Cannot duplicate INDArray");
  4.   if (this.isCompressed() && this.ordering() == Nd4j.order().charValue()) {
  5.     INDArray ret = Nd4j.createArrayFromShapeBuffer(data().dup(), this.shapeInfoDataBuffer());
  6.     ret.markAsCompressed(true);
  7.     return ret;
  8.   }
  9.   Nd4j.getCompressor().autoDecompress(this);
  10.   INDArray ret = Shape.toOffsetZeroCopy(this);
  11.   return ret;
  12. }

代码示例来源：origin: deeplearning4j/nd4j

  1. @Override
  2. public INDArray getrf(INDArray A) {
  3.   // FIXME: int cast
  4.   if (A.rows() > Integer.MAX_VALUE || A.columns() > Integer.MAX_VALUE)
  5.     throw new ND4JArraySizeException();
  6.   int m = (int) A.rows();
  7.   int n = (int) A.columns();
  8.   INDArray INFO = Nd4j.createArrayFromShapeBuffer(Nd4j.getDataBufferFactory().createInt(1),
  9.           Nd4j.getShapeInfoProvider().createShapeInformation(new int[] {1, 1}).getFirst());
  10.   int mn = Math.min(m, n);
  11.   INDArray IPIV = Nd4j.createArrayFromShapeBuffer(Nd4j.getDataBufferFactory().createInt(mn),
  12.           Nd4j.getShapeInfoProvider().createShapeInformation(new int[] {1, mn}).getFirst());
  13.   if (A.data().dataType() == DataBuffer.Type.DOUBLE)
  14.     dgetrf(m, n, A, IPIV, INFO);
  15.   else if (A.data().dataType() == DataBuffer.Type.FLOAT)
  16.     sgetrf(m, n, A, IPIV, INFO);
  17.   else
  18.     throw new UnsupportedOperationException();
  19.   if (INFO.getInt(0) < 0) {
  20.     throw new Error("Parameter #" + INFO.getInt(0) + " to getrf() was not valid");
  21.   } else if (INFO.getInt(0) > 0) {
  22.     log.warn("The matrix is singular - cannot be used for inverse op. Check L matrix at row " + INFO.getInt(0));
  23.   }
  24.   return IPIV;
  25. }

代码示例来源：origin: deeplearning4j/nd4j

  1. @Override
  2. public INDArray dup(char order) {
  3.   WorkspaceUtils.assertValidArray(this, "Cannot duplicate INDArray");
  4.   if (this.isCompressed() && this.ordering() == order) {
  5.     INDArray ret = Nd4j.createArrayFromShapeBuffer(data().dup(), this.shapeInfoDataBuffer());
  6.     ret.markAsCompressed(true);
  7.     return ret;
  8.   }
  9.   Nd4j.getCompressor().autoDecompress(this);
  10.   return Shape.toOffsetZeroCopy(this, order);
  11. }

代码示例来源：origin: deeplearning4j/nd4j

  1. @Override
  2. public INDArray convertToFloats() {
  3.   if (data.dataType() == DataBuffer.Type.FLOAT)
  4.     return this;
  5.   val factory = Nd4j.getNDArrayFactory();
  6.   val buffer = Nd4j.createBuffer(new long[]{this.length()}, DataBuffer.Type.FLOAT);
  7.   factory.convertDataEx(convertType(data.dataType()), this.data().addressPointer(), DataBuffer.TypeEx.FLOAT, buffer.addressPointer(), buffer.length());
  8.   return Nd4j.createArrayFromShapeBuffer(buffer, this.shapeInformation);
  9. }

代码示例来源：origin: deeplearning4j/nd4j

  1. @Override
  2. public void potrf(INDArray A, boolean lower) {
  3.   // FIXME: int cast
  4.   if (A.columns() > Integer.MAX_VALUE)
  5.     throw new ND4JArraySizeException();
  6.   byte uplo = (byte) (lower ? 'L' : 'U'); // upper or lower part of the factor desired ?
  7.   int n = (int) A.columns();
  8.   INDArray INFO = Nd4j.createArrayFromShapeBuffer(Nd4j.getDataBufferFactory().createInt(1),
  9.           Nd4j.getShapeInfoProvider().createShapeInformation(new int[] {1, 1}).getFirst());
  10.   if (A.data().dataType() == DataBuffer.Type.DOUBLE)
  11.     dpotrf(uplo, n, A, INFO);
  12.   else if (A.data().dataType() == DataBuffer.Type.FLOAT)
  13.     spotrf(uplo, n, A, INFO);
  14.   else
  15.     throw new UnsupportedOperationException();
  16.   if (INFO.getInt(0) < 0) {
  17.     throw new Error("Parameter #" + INFO.getInt(0) + " to potrf() was not valid");
  18.   } else if (INFO.getInt(0) > 0) {
  19.     throw new Error("The matrix is not positive definite! (potrf fails @ order " + INFO.getInt(0) + ")");
  20.   }
  21.   return;
  22. }

代码示例来源：origin: deeplearning4j/nd4j

  1. @Override
  2. public INDArray convertToDoubles() {
  3.   if (data.dataType() == DataBuffer.Type.DOUBLE)
  4.     return this;
  5.   val factory = Nd4j.getNDArrayFactory();
  6.   val buffer = Nd4j.createBuffer(new long[]{this.length()}, DataBuffer.Type.DOUBLE);
  7.   factory.convertDataEx(convertType(data.dataType()), this.data().addressPointer(), DataBuffer.TypeEx.DOUBLE, buffer.addressPointer(), buffer.length());
  8.   return Nd4j.createArrayFromShapeBuffer(buffer, this.shapeInformation);
  9. }

代码示例来源：origin: deeplearning4j/nd4j

  1. @Override
  2. public void gesvd(INDArray A, INDArray S, INDArray U, INDArray VT) {
  3.   // FIXME: int cast
  4.   if (A.rows() > Integer.MAX_VALUE || A.columns() > Integer.MAX_VALUE)
  5.     throw new ND4JArraySizeException();
  6.   int m = (int) A.rows();
  7.   int n = (int) A.columns();
  8.   byte jobu = (byte) (U == null ? 'N' : 'A');
  9.   byte jobvt = (byte) (VT == null ? 'N' : 'A');
  10.   INDArray INFO = Nd4j.createArrayFromShapeBuffer(Nd4j.getDataBufferFactory().createInt(1),
  11.           Nd4j.getShapeInfoProvider().createShapeInformation(new int[] {1, 1}).getFirst());
  12.   if (A.data().dataType() == DataBuffer.Type.DOUBLE)
  13.     dgesvd(jobu, jobvt, m, n, A, S, U, VT, INFO);
  14.   else if (A.data().dataType() == DataBuffer.Type.FLOAT)
  15.     sgesvd(jobu, jobvt, m, n, A, S, U, VT, INFO);
  16.   else
  17.     throw new UnsupportedOperationException();
  18.   if (INFO.getInt(0) < 0) {
  19.     throw new Error("Parameter #" + INFO.getInt(0) + " to gesvd() was not valid");
  20.   } else if (INFO.getInt(0) > 0) {
  21.     log.warn("The matrix contains singular elements. Check S matrix at row " + INFO.getInt(0));
  22.   }
  23. }

代码示例来源：origin: deeplearning4j/nd4j

  1. @Override
  2. public void geqrf(INDArray A, INDArray R) {
  3.   // FIXME: int cast
  4.   if (A.rows() > Integer.MAX_VALUE || A.columns() > Integer.MAX_VALUE)
  5.     throw new ND4JArraySizeException();
  6.   int m = (int) A.rows();
  7.   int n = (int) A.columns();
  8.   INDArray INFO = Nd4j.createArrayFromShapeBuffer(Nd4j.getDataBufferFactory().createInt(1),
  9.           Nd4j.getShapeInfoProvider().createShapeInformation(new int[] {1, 1}).getFirst());
  10.   if (R.rows() != A.columns() || R.columns() != A.columns()) {
  11.     throw new Error("geqrf: R must be N x N (n = columns in A)");
  12.   }
  13.   if (A.data().dataType() == DataBuffer.Type.DOUBLE) {
  14.     dgeqrf(m, n, A, R, INFO);
  15.   } else if (A.data().dataType() == DataBuffer.Type.FLOAT) {
  16.     sgeqrf(m, n, A, R, INFO);
  17.   } else {
  18.     throw new UnsupportedOperationException();
  19.   }
  20.   if (INFO.getInt(0) < 0) {
  21.     throw new Error("Parameter #" + INFO.getInt(0) + " to getrf() was not valid");
  22.   }
  23. }

代码示例来源：origin: deeplearning4j/nd4j

  1. Nd4j.getMemoryManager().memcpy(buffer, this.data());
  2.   copy = Nd4j.createArrayFromShapeBuffer(buffer, this.shapeInfoDataBuffer());
  3. } else {
  4.   copy = this.dup(this.ordering());

代码示例来源：origin: deeplearning4j/nd4j

  1. return Nd4j.createArrayFromShapeBuffer(buffer, this.shapeInfoDataBuffer());
  2. } else {
  3.   INDArray copy = Nd4j.createUninitialized(this.shape(), this.ordering());
  4.   copy = Nd4j.createArrayFromShapeBuffer(buffer, this.shapeInfoDataBuffer()); //this.dup(this.ordering());

代码示例来源：origin: deeplearning4j/nd4j

  1. Nd4j.getMemoryManager().memcpy(buffer, this.data());
  2.   copy = Nd4j.createArrayFromShapeBuffer(buffer, this.shapeInfoDataBuffer());
  3. } else {
  4.   copy = this.dup(this.ordering());

代码示例来源：origin: deeplearning4j/nd4j

  1. Nd4j.getMemoryManager().memcpy(buffer, this.data());
  2.   copy = Nd4j.createArrayFromShapeBuffer(buffer, this.shapeInfoDataBuffer());
  3. } else {
  4.   copy = this.dup(this.ordering());

代码示例来源：origin: org.nd4j/nd4j-api

  1. @Override
  2. public INDArray bitmapEncode(INDArray indArray, double threshold) {
  3.   DataBuffer buffer = Nd4j.getDataBufferFactory().createInt(indArray.length() / 16 + 5);
  4.   INDArray ret = Nd4j.createArrayFromShapeBuffer(buffer, indArray.shapeInfoDataBuffer());
  5.   bitmapEncode(indArray, ret, threshold);
  6.   return ret;
  7. }

代码示例来源：origin: org.nd4j/nd4j-cuda-10.0

  1. @Override
  2. public INDArray compress(INDArray array) {
  3.   //logger.info("Threshold [{}] compression", threshold);
  4.   Nd4j.getExecutioner().commit();
  5.   //Nd4j.getAffinityManager().ensureLocation(array, AffinityManager.Location.HOST);
  6.   DataBuffer buffer = compress(array.data());
  7.   if (buffer == null)
  8.     return null;
  9.   INDArray dup = Nd4j.createArrayFromShapeBuffer(buffer, array.shapeInfoDataBuffer());
  10.   dup.markAsCompressed(true);
  11.   return dup;
  12. }

内容来源于网络，如有侵权，请联系作者删除！

Nd4j

相关文章

    查看更多

    热门标签

    更多
    JavaquerypythonNode开发语言requestUtil数据库Table后端算法LoggerMessageElementParser

    最新文章

    更多

    Nd4j类方法

    • 蜀ICP备13028337号-1 大数据知识库 https://www.saoniuhuo.com © All rights reserved
    • 本站内容来源互联网,如果侵犯您的权益请联系我们删除, 联系方式：448109455@qq.com